Amazon CEO Andy Jassy could observe how profoundly artificial intelligence was transforming e-commerce.
In June, he informed employees that AI assistants would soon begin to permeate everyday life, “from shopping to travel to daily chores and tasks.”
Four months afterwards, Jassy mentioned during an earnings call that Amazon aims to collaborate with third-party agents and has been in discussions with several providers, although he did not disclose names.
Currently, Amazon is seeking to recruit a head of corporate development to assist in establishing strategic alliances in domains such as “agentic commerce,” as indicated in a recent job listing.
Amazon’s swift shift in its perspective on AI-driven commerce highlights how rapidly online retail is evolving, and the dangers the company faces if it doesn’t take decisive steps to retain oversight of its future.
The firm has observed how OpenAI, Google, Perplexity and Microsoft have introduced a surge of e-commerce assistants recently aimed at changing the shopping experience. Instead of going to Amazon, Walmart or Nike directly, shoppers could depend on AI agents to perform the tedious task of searching the web for the best price or ideal product, then purchase the item without leaving a chatbot interface.
The initial shopping agents from AI innovators appeared about a year ago. Consulting firm McKinsey estimated that agentic commerce could yield $1 trillion in U.S. retail revenue by 2030.
This trend presents a challenge to Amazon’s profit margins and customer relations. When a customer employs ChatGPT to initiate a transaction, for instance, OpenAI charges “a small fee” for each sale.
“With an agent on ChatGPT, retailers risk losing transactions on their platform to pay a toll on someone else’s highway for the same transaction,” Sucharita Kodali, a retail analyst at Forrester, remarked in a conversation.
Some firms are seeking to find a balance between collaborating with agent providers and competing against them. Walmart, Shopify and others have embraced a ‘frenemy’ strategy, announcing collaborations with AI firms while simultaneously enhancing their own tools and establishing boundaries on how agents may access their platforms.
Shopify CEO Tobi Lutke stated in a post on X on Tuesday that his company is “constructing all the layers of infrastructure to facilitate a new cambrian explosion of creativity in shopping.”
“I am truly thrilled about Agentic Commerce,” Lutke expressed. “There’s so much remarkable innovation happening. Everything I try feels delightful and right.”
Amazon has predominantly been in a defensive posture.
The company recently updated the underlying code of its website to prevent external AI agents from crawling it, as part of an initiative to protect its valuable training data from competitors. As of Tuesday, Amazon had barred 47 bots, including those from all the major AI firms, according to its website.
Amazon has even taken legal action. In November, Amazon sued Perplexity over an assistant in the startup’s Comet browser that permits it to make purchases on a user’s behalf. The company contended that Perplexity attempted to “conceal” its agents to continue scraping Amazon’s website without authorization.
Perplexity labeled the lawsuit a “bully tactic.”
Meanwhile, Amazon is investing substantially in its own AI technologies. The company introduced a shopping chatbot named Rufus last February, and has been evaluating an assistant called Buy For Me, capable of purchasing goods from other sites directly within Amazon’s e-commerce application.
Personalized shoppers
Morgan Stanley anticipates that by 2030, nearly half of American shoppers will utilize AI agents, and this technology could contribute up to $115 billion in U.S. e-commerce spending.
“We believe agentic commerce — essentially the capability to have a personal interactive digital shopper — is poised to be the next significant GenAI-enabled breakthrough,” Morgan Stanley analysts indicated in a November report.
They observed that a modest single-digit percentage of consumers currently initiate their “purchase journey” through AI, yet this could rise over time as around 40% to 50% of Americans presently use AI for product research.
Traffic directed from AI chatbots to U.S. retail websites has surged in recent months, particularly during the holiday season, but studies indicate that Google searches still outperform in terms of conversion rates and revenue per session.
AI-driven shopping remains an emerging market.
OpenAI’s Instant Checkout feature, introduced in ChatGPT in September, is only available for select products sold by Walmart, Shopify, Target and Etsy. Customers can only buy one item at a time, and they can’t link loyalty memberships like Walmart+.
Agents are also susceptible to errors.
Scot Wingo, founder of e-commerce software firm ReFiBuy, recently tried Perplexity’s Instant Buy feature that enables users to shop directly within its search engine.
Wingo attempted to buy a cable-knit sweater from Abercrombie & Fitch, but Perplexity’s agent continuously displayed error messages, even though both items were available on the retailer’s site. He ultimately abandoned the effort.
Earlier this month, Wingo was searching for a coffee maker on ChatGPT when it suggested a Breville espresso machine. Upon clicking the product, he was taken aback to see an image of a garden rake.
“These crawlers obtain this data, and you never know precisely what they will retrieve,” Wingo mentioned.
‘Leader’s dilemma’
As Amazon considers its forthcoming strategy regarding shopping agents, it is discreetly permitting them to access certain of its properties.
Subsidiaries such as shoe retailer Zappos, fashion website Shopbop and deals site Woot do not seem to have any provisions restricting agents in their robots.txt files, which specify how crawlers can explore particular web pages.
“Often, they will utilize the subsidiaries for experimentation,” Wingo noted. “Zappos possesses its own experience and database, so it’s not as if they’re completely unleashing everything.”
The company could ultimately adopt strategies employed by its competitors if it decides to allow agents access to its main e-commerce platform. Shopify and Walmart have established parameters around what external shopping agents are permitted to do on their platforms.
Amazon may be open to allowing agents access to its catalog, but it likely aims to safeguard more critical data from its rivals, Wingo stated, including its extensive collection of customer reviews and sales rankings, both of which indicate a product’s quality and can enhance an AI chatbot’s responses.
“Those are probably the two most proprietary data points that, if I’m Amazon, I want to secure,” Wingo remarked.
Amazon isn’t abandoning its in-house tools.
Rufus’ functionality has improved since Amazon initially launched it last year, and the company has been promoting the chatbot across more segments of its site to boost user adoption.
Amazon recently introduced a feature that enables Rufus to automatically purchase items on behalf of a Prime user once they reach a certain price. The chatbot now also recommends products from various websites, not solely from Amazon.
In recent weeks, Amazon has begun experimenting with a feature that allows Rufus to develop custom shopping guides, akin to OpenAI’s “shopping research” tool launched last month.
“Instead of the innovator’s dilemma, I would classify Amazon as facing what I term the leader’s dilemma,” stated Jordan Berke, founder and CEO of retail consulting firm Tomorrow. “Their market presence is so substantial that they stand to lose the most.”
